US 12,259,921 B1
Systems and methods for clustering digital content
Jennifer Rose Guriel, Kirkland, WA (US); Abdelrahman Khaled Abdo Mohamed, Seattle, WA (US); Mingqi Wu, Bellevue, WA (US); Gershom Payzer, Redmond, WA (US); Christopher Ian Charla, Seattle, WA (US); Madeline Jaye Whisenant, Seattle, WA (US); Bridgette Marie Kuehn, Seattle, WA (US); Jianxun Lian, Beijing (CN); and Licheng Pan, Hangzhou (CN)
Assigned to Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Dec. 5, 2023, as Appl. No. 18/529,975.
Int. Cl. G06F 16/45 (2019.01); G06F 40/40 (2020.01)
CPC G06F 16/45 (2019.01) [G06F 40/40 (2020.01)] 20 Claims
OG exemplary drawing
 
1. A method of generating a cluster description for a grouping of interactive multimedia content items, comprising:
identifying a collection of interactive multimedia content items to be characterized according to a set of candidate attribute tags;
identifying usage data associated with engagement of a plurality of users with the collection of interactive multimedia content items;
for each interactive multimedia content item of the collection of interactive multimedia content items:
providing a first prompt as a first input to a large language model (LLM), the first prompt including a title of the interactive multimedia content item and the set of candidate attribute tags; and
receiving, from the LLM and based on the first prompt, a first output including a subset of the candidate attribute tags determined to be associated with the interactive multimedia content item;
providing a second prompt as a second input to the LLM, the second prompt including titles of interactive multimedia content items of a cluster of interactive multimedia content items determined to be related, the second prompt further including candidate attribute tags associated with each of the titles of the interactive multimedia content items of the cluster of interactive multimedia content items; and
receiving, from the LLM, a second output including a cluster title and a cluster description associated with the cluster of interactive multimedia content items, the second output being generated by the LLM based candidate attribute tags associated with the titles of interactive multimedia content items of the cluster.